A Predictive Diagnostic System with Self-Learning for Space Mechanisms

SpaceWorks and the Georgia Institute of Technology propose an integrated predictive diagnostic system for evaluating the condition of spacecraft moving parts. Specific mechanisms to be addressed include reaction and momentum wheels, gimbals, flywheels and cryocoolers. To accurately estimate the remaining life of components, an intelligent diagnostic and prognostic monitoring and computational strategy is proposed, capable of using sensor data with adaptive and stochastic algorithms employing self-learning capability and statistical confidence. It is envisioned that the failure lifetime of the application will be asymptotically realized in real-time. Given that prognostic techniques have not been well developed to address uncertainties in defect growth behavior, the proposed methodology will use built-in computational capability to instantly update the predictive algorithms for reliable health assessment and reporting. Our investigation will focus on low-mass sensors and miniaturized electronics for on-board data acquisition, storage and processing.